AI Video Cost per Second 2026: Pricing Calculator

Q: How much does AI video cost per second in 2026?

There is no single number — it depends on model and resolution. At regular list rates, Google's Gemini Omni 1.1 Flash is the floor at $0.03 per second for 360p draft output, $0.10/s at 720p, $0.15/s at 1080p and $0.30/s at 4K. fal MiniMax H3 Max's $0.025/s at 480p is a temporary 50%-off promo through Sep 30, 2026 (768p promo $0.04/s, 1080p promo $0.08/s; then $0.05/$0.08/$0.16 list). Other regular list rates run $0.05–$0.70/s: Runway Gen-4 Turbo $0.05/s, Veo 3.1 Lite $0.05/s, Sora 2 $0.10/s, Runway Gen-4.5 $0.12/s, Kling $0.084–$0.42/s, Sora 2 Pro $0.30–$0.70/s.

Q: What is the cheapest AI video generator per second in 2026?

Gemini Omni 1.1 Flash (Google, GA Aug 27, 2026) has the cheapest regular list per-second API rate among major vendors: $0.03 per second for its 360p draft tier — roughly one-third the cost of its own standard 720p output. Runway Gen-4 Turbo ($0.05/s) and Veo 3.1 Lite ($0.05/s) are next. fal MiniMax H3 Max's $0.025/s at 480p undercuts them while it lasts, but it is a 50%-off promo ending Sep 30, 2026 — after that H3 Max lists $0.05/s at 480p, $0.08/s at 768p and $0.16/s at 1080p, with a free tier of 5 clips/day.

Q: How much does Gemini Omni 1.1 Flash cost per second?

Gemini Omni 1.1 Flash bills per second of output video by resolution: $0.03/s at 360p (draft mode), $0.10/s at 720p (standard), $0.15/s at 1080p and $0.30/s at 4K. 1080p and 4K are upscaled from 720p, not native renders. A 10-second 360p draft costs about $0.30; a 10-second 1080p clip costs about $1.50.

Q: How does Gemini Omni 1.1 Flash compare to Sora and Runway on price?

Gemini Omni 1.1 Flash starts at $0.03/s (360p draft) and matches Sora 2 at $0.10/s for 720p, with 1080p at $0.15/s and 4K at $0.30/s. Sora 2 is $0.10/s at 720p and its API sunsets September 24, 2026; Sora 2 Pro is $0.30–$0.70/s. Runway Gen-4 Turbo is $0.05/s and Gen-4.5 is $0.12/s (both credit-based, $0.01/credit). Omni adds 40-second scenes, keyframe control and native audio that per-second video SKUs from Veo don't include.

Q: What does a 30/60/90-second AI video cost with Gemini Omni 1.1 Flash?

At the 360p draft tier ($0.03/s): $0.90 for 30 seconds, $1.80 for 60 seconds, $2.70 for 90 seconds. At standard 720p ($0.10/s): $3.00 / $6.00 / $9.00. At 1080p ($0.15/s): $4.50 / $9.00 / $13.50. At 4K ($0.30/s): $9.00 / $18.00 / $27.00. Use the calculator above for any model and duration.

Q: What does a perpetual AI live stream (24/7 video) cost per month?

Run a video model continuously and the bill is per-second math times 86,400 seconds/day. At fal MiniMax H3 Max's 480p promo rate of $0.025/s (50%-off through Sep 30, 2026), a 24/7 stream costs ~$65.8K/mo: $0.025 × 86,400 s/day × 30.44 days = $65,750.40. At the $0.05/s 480p list rate fal lists for after Sep 30, the same stream runs ~$131.5K/mo. For comparison, levelsio (Pieter Levels) put the stream he runs on his Infinite Slop TV at $32,940/mo in his Sep 2, 2026 post, when the promo then in force was 75% off, against ~$15K/mo ad revenue — profitable only once more ad slots sell (@levelsio on X).

Published August 30, 2026 · Updated September 20, 2026By ABD Legacy LLC

AI video cost per second 2026 Gemini Omni 1.1 Flash pricing Video API cost comparison Higgsfield + GPT-6 Astra ad cost

The short answer: AI video in 2026 bills per second of output, and the new price floor is Gemini Omni 1.1 Flash (Google) at $0.03 per second for its 360p draft tier — with $0.10/s at 720p, $0.15/s at 1080p and $0.30/s at 4K. On that ladder a 30-second video costs $0.90, a 60-second video $1.80, and a 90-second video $2.70 (draft quality). The calculator below runs the math live for every major vendor — Gemini Omni, Sora, Runway, Kling, Veo and MiniMax H3 Max — so you can quote video work from a real number instead of a guess.

This page is the pricing companion to "AI Video Generation Cost 2026: Gemini Omni vs Sora Runway" (Find AI Agency) — the strategy is there; the per-second math is here. Google GA'd Gemini Omni 1.1 Flash on Aug 27, 2026 as the stable successor to the June preview, and its per-second pricing ladder lands it directly against Sora 2, Runway and Kling in the video price war.

Video Cost per Second Calculator

Model (per-second API rate; fal H3 Max promo re-read Sep 20, 2026)

Resolution / tier

Video length (seconds)

Quick duration

Estimated cost for this video—

Effective price per second—

Cost per minute (60s)—

Model—

Vendor / released—

Same model, standard durations

30-second video—

60-second video—

90-second video—

All figures are published list rates as of Sep 2, 2026 except the fal MiniMax H3 Max promo row, re-read Sep 20, 2026 (sources below). The Gemini Omni 1.1 Flash $0.03/s headline is its 360p draft tier; standard 720p output is $0.10/s. Sora 2 API sunsets Sept 24, 2026. fal MiniMax H3 Max is on 50%-off promo pricing ($0.025/s at 480p, $0.04/s at 768p, $0.08/s at 1080p) through Sep 30, 2026 as published on fal's model pages (re-read Sep 20, 2026), then reverts to $0.05/s (480p), $0.08/s (768p) and $0.16/s (1080p). Re-verify provider rates before quoting a client — 2026 pricing moves weekly.

Perpetual-stream example (Sep 20, 2026): run MiniMax H3 Max continuously at fal.ai's current 480p promo rate of $0.025/s and a month of 24/7 output costs ~$65.8K — $0.025 × 86,400 s/day × 30.44 days = $65,750.40. That is the 50%-off rate fal publishes for the promo it lists as ending Sep 30, 2026. Under the steeper 75%-off promo in force when levelsio's Infinite Slop TV post went up on Sep 2, 2026, the same run was priced at $32,940/mo against ~$15K/mo ad revenue — profitable only after more ad slots sell (x.com/levelsio). The promo ends Sep 30, 2026; at the $0.05/s 480p list rate the same 24/7 stream runs ~$131.5K/mo.

How the math works

Every major video API bills per second of output video, by resolution. The total is a single multiplication: duration (seconds) × per-second price (tier). That is the whole formula — the tiers are what differ.

Cost per video = durationSeconds × pricePerSecond(tier) Example — Gemini Omni 1.1 Flash, 360p draft tier ($0.03/s): 30s → 30 × $0.03 = $0.90 60s → 60 × $0.03 = $1.80 90s → 90 × $0.03 = $2.70 Example — Gemini Omni 1.1 Flash, standard 720p ($0.10/s): 30s → 30 × $0.10 = $3.00 60s → 60 × $0.10 = $6.00 90s → 90 × $0.10 = $9.00 Example — fal MiniMax H3 Max, 480p promo ($0.025/s, thru Sep 30 2026): 60s → 60 × $0.025 = $1.50/min (After Sep 30 the 480p list rate is $0.05/s → $3.00/min; 768p list is $0.08/s → $4.80/min; 1080p list is $0.16/s → $9.60/min.)

The Google Cloud pricing page confirms the Omni ladder with token math: 1,931 tokens/second at 360p, 5,792 at 720p, 8,688 at 1080p, 17,376 at 4K, billed at $17.50 per 1M video-output tokens — which works out to ≈$0.034/s, ≈$0.101/s, ≈$0.152/s and ≈$0.304/s respectively, rounded to the published $0.03/$0.10/$0.15/$0.30 ladder.

Per-second video pricing table (Sep 2026)

Current as of Sep 2, 2026, except the fal MiniMax H3 Max promo row, re-read Sep 20, 2026. The Gemini Omni 1.1 Flash row is highlighted — it went GA Aug 27, 2026 and sets the new per-second floor at standard list rates. fal MiniMax H3 Max's 50%-off promo ($0.025/s 480p, $0.04/s 768p, $0.08/s 1080p through Sep 30, 2026) is listed separately as the promo tier.

Model (vendor)	Price per second	Released	Notes
Gemini Omni 1.1 Flash NEW Google	$0.03 (360p draft) · $0.10 (720p) · $0.15 (1080p) · $0.30 (4K)	Aug 27, 2026	40s scenes (10s extensions), first/last frame control, native audio; 1080p/4K upscaled from 720p. blog.google · fal.ai
Veo 3.1 Lite Google	$0.05 (720p) · $0.08 (1080p)	—	No 4K tier
Veo 3.1 Fast Google	$0.10 (720p) · $0.12 (1080p) · $0.30 (4K)	—	Google's fidelity line
Sora 2 OpenAI	$0.10 (720p)	—	API sunsets Sept 24, 2026
Sora 2 Pro OpenAI	$0.30 (720p) · $0.70 (1080p)	—	Batch = half price
Runway Gen-4 Turbo Runway	$0.05 (5 cr/s)	—	Credit-based; $0.01/credit API
Runway Gen-4.5 Runway	$0.12 (12 cr/s)	—	Credit-based; $0.01/credit API
Kling API Kuaishou	$0.084 (720p, no audio) · $0.112 (720p, native audio) · $0.14 (1080p) · $0.42 (4K)	—	Pricing varies with resolution and audio
fal MiniMax H3 Max 50% promo thru Sep 30 fal.ai	$0.025 (480p promo) · $0.04 (768p promo) · $0.08 (1080p promo) — thru Sep 30, 2026; then $0.05 (480p) · $0.08 (768p) · $0.16 (1080p) list	—	fal.ai promo pricing (50% off) through Sep 30, 2026, then reverts to $0.05/s (480p), $0.08/s (768p) and $0.16/s (1080p) list; free 5 clips/day; 5s clip renders <3s
Higgsfield + GPT-6 Astra Paid Ads plugin NEW Higgsfield × OpenAI	$0.16 (modeled, per second of finished ad: generation $0.09 + Astra tokens $0.07 + Effects $0.00) modeled	Bundle Sept 18, 2026	Ad-production stack, not a raw API rate — inputs and assumptions in the section below. Generation priced at Higgsfield's published API rates, orchestration at OpenAI's published GPT-6 Astra token rates, Higgsfield Effects generations free. The ChatGPT plugin itself predates the bundle date. higgsfield.ai

Higgsfield + GPT-6 Astra: what an AI ad costs per second

Higgsfield's “Paid Ads” bundle for GPT-6 Astra is live in ChatGPT today. The vendor's own hero card reads “Higgsfield x GPT-6 Astra for Paid Ads: Market research to your next ad test - now on ChatGPT” and links to the Paid Ads bundle of eight workflows — Campaign Analyst, Ad Strategist, Customer Voice Ads, Headline Multiplier, Ad Resizer, Ad Localizer, Ad Recreator and Hook Multiplier — each one invoked from ChatGPT as /marketing. The bundle's dated vendor announcement is September 18, 2026 (@higgsfield on X). Note what that date covers: the bundle, not the ChatGPT plugin — the plugin predates it, and Higgsfield publishes no launch date for the plugin itself.

Ad production, per second of finished video	Higgsfield + GPT-6 Astra plugin	Video editor + media buyer (incumbent)
Generation / production	$0.09 — 3 takes × 30s = 90 generated seconds at Higgsfield's published Wan 3.0 API rate of $0.03/s (40% off; $0.05/s list)	$10.00 — 4.0 h of editing at $75/h ÷ 30 s assumed
Research, hooks, campaign ops	$0.07 — 120K input + 18K output GPT-6 Astra tokens = $2.10 at $10 / $50 per 1M	$5.00 — 2.0 h of media buying at $75/h ÷ 30 s assumed
Effects / viral presets	$0.00 — Higgsfield Effects, “Viral video presets now in ChatGPT, with free generations”	—
Total per second	$0.16 / second ($4.80 for a 30-second ad)	$15.00 / second ($450 for a 30-second ad)
Delta per second	$14.84 cheaper per second — roughly 94× on these inputs

Inputs and assumptions. Sourced (Sep 20, 2026): Higgsfield's published API per-second generation rates from the vendor's own model catalog (open.higgsfield.ai, reached from the API link on higgsfield.ai) — Wan 3.0 from $0.03/s at 40% off (list $0.05/s) and Higgsfield's own Cinema Studio 4.0 from $0.2057/s; “free generations” for Higgsfield Effects presets in ChatGPT (higgsfield.ai); and OpenAI's published GPT-6 Astra token rates of $10 per 1M input / $50 per 1M output under the 272K threshold (the full ladder is on GPT-6 Astra API pricing). Assumed, not sourced: a 3.0× take ratio (three generated takes per finished 30-second ad), 120K input + 18K output Astra tokens per ad across three workflow runs, and a six-hour video editor + media buyer pipeline at $75/hour blended. Higgsfield publishes no per-generation price for the ChatGPT plugin itself, so the generation line is priced at its published API rates instead. Raise either assumption and the plugin figure moves with it: at the $0.05/s list rate the same model is $0.22/s, and at Cinema Studio 4.0's $0.2057/s it is $0.69/s — still an order of magnitude under the assumed human pipeline.

Where this fits for an agency: the plugin moves the cost of producing the ad, not the decision about what to test. The same stack's computer-use side — what Astra automating the rest of the delivery changes — is on Computer-Use Agents Change What Agencies Can Automate.

What 30 / 60 / 90-second videos cost (worked examples)

Straight multiplication on the table above. These are raw API generation costs — before agency markup, retries, or client pricing.

Gemini Omni 1.1 Flash (360p draft, $0.03/s): 30s = $0.90 · 60s = $1.80 · 90s = $2.70
Gemini Omni 1.1 Flash (720p, $0.10/s): 30s = $3.00 · 60s = $6.00 · 90s = $9.00
Gemini Omni 1.1 Flash (1080p, $0.15/s): 30s = $4.50 · 60s = $9.00 · 90s = $13.50
Gemini Omni 1.1 Flash (4K, $0.30/s): 30s = $9.00 · 60s = $18.00 · 90s = $27.00
Runway Gen-4 Turbo ($0.05/s): 30s = $1.50 · 60s = $3.00 · 90s = $4.50
fal MiniMax H3 Max (480p promo thru Sep 30, 2026, $0.025/s): 30s = $0.75 · 60s = $1.50 · 90s = $2.25 — after Sep 30 the 480p list rate is $0.05/s (60s = $3.00)
Sora 2 (720p, $0.10/s): 30s = $3.00 · 60s = $6.00 · 90s = $9.00 — until the API sunset on Sept 24, 2026

Why Gemini Omni 1.1 Flash resets the per-second floor

Google GA'd Gemini Omni 1.1 Flash on Aug 27, 2026 — the stable successor to the Gemini Omni Flash preview (June 30, 2026) — available via the Gemini API in Google AI Studio, the Gemini Enterprise Agent Platform, Google Flow (AI Plus/Pro/Ultra), and scene extension in the Gemini app. Its 360p draft tier at $0.03/s (~60% faster and roughly one-third the cost of its own 720p output) is the cheapest listed per-second rate among major vendors; its $0.10/s 720p tier matches Sora 2, Veo 3.1 Fast and the previous Omni Flash.

What you get at that price: 40-second scenes built from 10-second extensions, first/last frame control, up to 3 seconds of video reference, 10 seconds of lookback context (up from 1s), and native synchronized audio. 1080p and 4K are upscaled outputs, not native renders — the standard generation tier is 720p. For agencies, the practical takeaway: a usable 30-second draft costs under a dollar, and the same clip at client-ready 720p costs $3.00 in raw API spend.

Watch-outs: Sora 2's API shuts down September 24, 2026 — any quote built on Sora per-second math has a hard expiry. And fal MiniMax H3 Max's attention-grabbing $0.025/s is a 50%-off promo through Sep 30, 2026; the list price then reverts to $0.05/s at 480p, $0.08/s at 768p and $0.16/s at 1080p — the same perpetual stream that costs ~$65.8K/mo during the promo runs ~$131.5K/mo at the $0.05/s 480p list rate.

Quote the whole project, not just the video

Open the AI Agency Pricing Calculator →

The full calculator covers setup fees, retainers, model strategy, failure/retry risk and margin — with current 2026 rates. For token-model costs see AI Model Cost per Task 2026.

Frequently asked questions

How much does AI video cost per second in 2026?

There is no single number — it depends on model and resolution. At regular list rates, Google's Gemini Omni 1.1 Flash is the floor at $0.03 per second for 360p draft output, $0.10/s at 720p, $0.15/s at 1080p and $0.30/s at 4K. fal MiniMax H3 Max's $0.025/s at 480p is a temporary 50%-off promo through Sep 30, 2026 (768p promo $0.04/s, 1080p promo $0.08/s; then $0.05/$0.08/$0.16 list). Other regular list rates run $0.05–$0.70/s: Runway Gen-4 Turbo $0.05/s, Veo 3.1 Lite $0.05/s, Sora 2 $0.10/s, Runway Gen-4.5 $0.12/s, Kling $0.084–$0.42/s, Sora 2 Pro $0.30–$0.70/s.

What is the cheapest AI video generator per second in 2026?

Gemini Omni 1.1 Flash (Google, GA Aug 27, 2026) has the cheapest regular list per-second API rate among major vendors: $0.03 per second for its 360p draft tier — roughly one-third the cost of its own standard 720p output. Runway Gen-4 Turbo ($0.05/s) and Veo 3.1 Lite ($0.05/s) are next. fal MiniMax H3 Max's $0.025/s at 480p undercuts them while it lasts, but it is a 50%-off promo ending Sep 30, 2026 — after that H3 Max lists $0.05/s at 480p, $0.08/s at 768p and $0.16/s at 1080p, with a free tier of 5 clips/day.

How much does Gemini Omni 1.1 Flash cost per second?

Gemini Omni 1.1 Flash bills per second of output video by resolution: $0.03/s at 360p (draft mode), $0.10/s at 720p (standard), $0.15/s at 1080p and $0.30/s at 4K. 1080p and 4K are upscaled from 720p, not native renders. A 10-second 360p draft costs about $0.30; a 10-second 1080p clip costs about $1.50.

How does Gemini Omni 1.1 Flash compare to Sora and Runway on price?

Gemini Omni 1.1 Flash starts at $0.03/s (360p draft) and matches Sora 2 at $0.10/s for 720p, with 1080p at $0.15/s and 4K at $0.30/s. Sora 2 is $0.10/s at 720p and its API sunsets September 24, 2026; Sora 2 Pro is $0.30–$0.70/s. Runway Gen-4 Turbo is $0.05/s and Gen-4.5 is $0.12/s (both credit-based, $0.01/credit). Omni adds 40-second scenes, keyframe control and native audio that per-second video SKUs from Veo don't include.

What does a 30/60/90-second AI video cost with Gemini Omni 1.1 Flash?

At the 360p draft tier ($0.03/s): $0.90 for 30 seconds, $1.80 for 60 seconds, $2.70 for 90 seconds. At standard 720p ($0.10/s): $3.00 / $6.00 / $9.00. At 1080p ($0.15/s): $4.50 / $9.00 / $13.50. At 4K ($0.30/s): $9.00 / $18.00 / $27.00. Use the calculator above for any model and duration.

What does a perpetual AI live stream (24/7 video) cost per month?

Run a video model continuously and the bill is per-second math times 86,400 seconds/day. At fal MiniMax H3 Max's 480p promo rate of $0.025/s (50%-off through Sep 30, 2026), a 24/7 stream costs ~$65.8K/mo: $0.025 × 86,400 s/day × 30.44 days = $65,750.40. At the $0.05/s 480p list rate fal lists for after Sep 30, the same stream runs ~$131.5K/mo. For comparison, levelsio (Pieter Levels) put the stream he runs on his Infinite Slop TV at $32,940/mo in his Sep 2, 2026 post, when the promo then in force was 75% off, against ~$15K/mo ad revenue — profitable only once more ad slots sell (@levelsio on X).

Sources

Google Blog, "Build with Gemini Omni 1.1 Flash" (Aug 27, 2026): blog.google
Google Cloud, Gemini Enterprise Agent Platform — Generative AI pricing (token math confirming the per-second ladder): cloud.google.com
fal.ai, Gemini Omni Flash 1.1 Text to Video (Google partner listing, per-second billing): fal.ai
The Decoder, "Google's Gemini Omni 1.1 Flash makes AI video generation cheaper and more flexible" (Aug 27, 2026): the-decoder.com
fal.ai, MiniMax H3 Max model pages — 50%-off promo $0.025/s (480p) / $0.04/s (768p) / $0.08/s (1080p) through Sep 30, 2026, then $0.05/$0.08/$0.16 list; re-read Sep 20, 2026; free 5/day: fal.ai — text-to-video · image-to-video
levelsio on X — "New price for Minimax H3 Max dropped… $0.0125/s*60s*60m*24h*30.5d=$32,940/mo… I'm at $15K/mo revenue now" (Sep 2, 2026): x.com/levelsio
Unifically, Sora API pricing + Sept 24, 2026 shutdown: unifically.com
CostBench, Kling API pricing: costbench.com
Melies, Runway ML pricing ($0.01/credit; Gen-4 Turbo 5 cr/s, Gen-4.5 12 cr/s): melies.co
Higgsfield, homepage hero + Higgsfield Effects card (live, undated) — “Higgsfield x GPT-6 Astra for Paid Ads: Market research to your next ad test - now on ChatGPT” and “Viral video presets now in ChatGPT, with free generations”: higgsfield.ai
Higgsfield, Paid Ads bundle — the eight GPT-6 Astra ad workflows, invoked as /marketing (live, undated): higgsfield.ai/gpt-astra/bundles/paid-ads-bundle
Higgsfield, API model catalog — the published per-second generation rates used in the ad-production model (Wan 3.0 from $0.03/s at 40% off; Cinema Studio 4.0 from $0.2057/s), read Sep 20, 2026: open.higgsfield.ai/explore
@higgsfield on X, “Meet Higgsfield x GPT-6 Astra for Paid Ads” (Sept 18, 2026) — the dated vendor announcement for the bundle: x.com/higgsfield

Accuracy note: All per-second prices are published list rates as of Sep 2, 2026 except the fal MiniMax H3 Max promo row, re-read Sep 20, 2026 (verified fact sheets t_39e6123c + t_d77cb9c2). Gemini Omni 1.1 Flash $0.03/s is the 360p draft tier — standard 720p is $0.10/s; 1080p/4K are upscaled, not native. Sora 2 API sunset (Sept 24, 2026) and fal MiniMax H3 Max's 50%-off promo (ends Sep 30, 2026; then $0.05/s 480p / $0.08/s 768p / $0.16/s 1080p list, re-read Sep 20, 2026) are dated facts — recompute after those dates. Speed/quality claims for Omni's draft mode are vendor-reported. Re-verify provider rates before quoting clients. Higgsfield + GPT-6 Astra figures were added Sep 20, 2026: the plugin-side costs are the vendor's published API per-second rates and OpenAI's published GPT-6 Astra token rates, the human "video editor + media buyer" comparison is a stated modeling assumption rather than a sourced rate, and the ChatGPT plugin itself carries no published launch date.